Determining Driving Risk Factors from Near-Miss Events in Telematics Data Using Histogram-Based Gradient Boosting Regressors

This study introduces a novel method for driving risk assessment based on the analysis of near-miss events captured in telematics data. Near-miss events, which are highly correlated with accidents, are employed as proxies for accident prediction. This research employs histogram-based gradient boosti...

ver descrição completa

Detalhes bibliográficos
Autores: Sun, Shuai, Guillén, Montserrat, Pérez Marín, Ana María, Ni, Linglin
Tipo de documento: artigo
Estado:Versão publicada
Data de publicação:2024
País:España
Recursos:Universidad de Barcelona
Repositório:Dipòsit Digital de la UB
OAI Identifier:oai:diposit.ub.edu:2445/219126
Acesso em linha:https://hdl.handle.net/2445/219126
Access Level:Acceso aberto
Palavra-chave:Assegurances d'automòbils
Risc (Assegurances)
Models lineals (Estadística)
Telemàtica
Automobile insurance
Risk (Insurance)
Linear models (Statistics)
Telematics
Descrição
Resumo:This study introduces a novel method for driving risk assessment based on the analysis of near-miss events captured in telematics data. Near-miss events, which are highly correlated with accidents, are employed as proxies for accident prediction. This research employs histogram-based gradient boosting regressors (HGBRs) for the analysis of telematics data, with comparisons made across datasets from China and Spain. The results presented in this paper demonstrate that HGBR outperforms conventional generalized linear models, such as Poisson regression and negative binomial regression, in predicting driving risks. Furthermore, the findings suggest that near-miss events could serve as a substitute for traditional claims in calculating insurance premiums. It can be seen that the machine learning algorithm offers the prospect of more accurate risk assessments and insurance pricing.